Can you introduce yourself and SLICES-RI (Research Infrastructures)?

I’m a senior researcher at INRIA working in the Avalon team in Lyon, and I’m co-coordinating SLICES-IP an European project aiming at setting up SLICES-RI. The goal of this European research infrastructure is to enable experimentation across the whole digital continuum. It involves 26 partners from up to sixteen countries and each country contributes with potentially different kinds of hardware and expertise along the digital continuum.

So the idea is to provide a global platform that enables researchers and industries to experiment either on a specific part of the digital continuum or to end-to-end experiment across the whole continuum. This makes it possible to test scenarios from IoT devices up to cloud or HPC systems, and also to evaluate the impact of technologies such as AI across the different layers of the infrastructure.

« Our goal is to help them understand how to design and size their infrastructures. »

What challenges share CERN and SKAO?

Even if they are very different scientific experiments, they both produce extremely large quantities of data, and both need to process these data through complex workflows. So they face similar questions: how to manage the data, how to manage the computation efficiently, and how to control performance, cost, and environmental impact.

Another important point is that these infrastructures are designed to operate for many years. However, hardware and software technologies evolve much faster than the lifetime of these scientific infrastructures. So one of the key challenges is how to upgrade both the hardware and the software over time. By experimenting with different possible evolutions of their infrastructure, we can help them plan and optimize future upgrades.

« We have deployed a benchmark suite developed for SKA to test and optimise their pipeline. »

So you provide both training and expertise….

Yes. The hardware and software stacks required to build infrastructures like LHCb or SKA are very complex and the experimental platforms used to study them also require specific skills. So one of the long-term goals is to make these infrastructures easier to use for engineers and researchers working in these scientific communities, and in particular to challenges such a measuring the energy consumption or the environmental impact.

How does this work benefit the SLICES infrastructure itself?

SLICES-RI is still a relatively young research infrastructure so the ODISSEE project is interesting for us because it allows us to work with large scientific infrastructures that have very demanding requirements in terms of data, computing, networking, and storage. This allows us to validate some of the design choices of the platform, and potentially identify improvements.

One example is the development of domain-specific services. For instance, the benchmark suite for SKA allows us to study how we can provide services tailored to specific research communities. The idea is to provide interfaces that are adapted to the needs of these communities, while minimizing the operational cost for the infrastructure and avoiding unnecessary complexity for the users.